United Kingdom

Demographic Horizons™

Mortality Improvements

Aon’s mortality improvements offering includes:

  • Robust improvements model for scheme-specific profiling based on members’ affluence (re-expressed in terms of industry-standard CMI Model (‘A’ parameter) for ease of use)
  • Adjustment factors to apply to historic mortality improvements by age, sex and month, to allow pandemic experience to be used in mortality analyses
  • A post-pandemic view on longevity improvement that we review quarterly to allow for shifting industry consensus, evolving views (in absence of predictive data from population-level mortality), and emerging post-pandemic data

Mortality improvements

Mortality improvements are the rates at which mortality rates are projected to change in the future. Even the largest pension schemes do not have sufficient data to determine their own mortality improvement trends, and so it is necessary to look at wider data. The starting point is usually a smoothed version of the improvements seen in national (e.g. England & Wales) population data, with very broad adjustments for the scheme’s socio-economic profile. The industry standard is the model produced annually by the CMI, which smooths national-level improvements by sex, age and year of birth, and allows users to define their own adjustments to apply to this.

‘A’ parameter analysis

Mortality improvements over the past two decades have varied by socioeconomic group (SEG), with higher SEGs having seen higher improvement rates.

Our ‘A’ parameter analysis involves

  • capturing a pension scheme’s profile using postcode data;
  • benchmarking it against
    • the England & Wales population and pension schemes in the Demographic Horizons dataset;
  • performing a tailored calibration of the CMI model which allows for the scheme’s profile.

This is a robust method for capturing scheme specifics. In contrast, other approaches to project separate improvement trends by subgroup have proved brittle.

Sub-annual mortality adjustments

Mortality experience, particularly in 2020 and 2021, was exceptionally heavy due to the impact of COVID-19. There are three possible approaches to handling this data in a mortality experience analysis:

  1. Use the exceptional data unadjusted - The unadjusted mortality is generally viewed as not being predictive of future mortality. Accordingly, this approach is a non-starter.
  2. Ignore the exceptional data entirely - This is the simplest choice. The primary concern is that it discards potentially informative data. In addition, if the heavy 2020 and 2021 years are excluded but there are knock-on effects of the pandemic, for example a frailty effect leading to lower mortality in say 2022 then this is inconsistent and potentially biases predicted future mortality.
  3. Adjust the data in some way so that it has predictive value - This is the approach we are adopting for our own mortality experience analytics. In our view, including the mortality experience from 2020 and 2021 with suitable adjustments in conjunction with several years of pre and post pandemic mortality experience is, on balance, likely to be a more robust approach than simply excluding 2020 / 2021 and relying on just the pre-pandemic years.

Based on analysis of population-level data, Aon’s approach is to adjust mortality experience to allow for the population-level impact of COVID-19. The data available suggest that:

  • Adjustment factors need to vary at least monthly to capture the varying level of excess mortality
  • Geographic variation in excess mortality means that location needs to be allowed for (in some years)
  • Simple multiplicative factors are sufficient to capture mortality variation across the socio-economic spectrum

Post-pandemic view

The latest version of the industry-standard mortality projections model in the UK (CMI_2021, released in March 2022) places nil weight on the last two years of data.

We continue to express our expectations for mortality and set our best estimates for mortality improvement based on the latest version of the CMI model, but we do expect that, for the next few years at least, extrapolative models will struggle to provide a suitable and consistent forward-looking mortality projection, because it is hard to extract a meaningful trend from years containing (and following) pandemic mortality.

Monitoring the consensus

Following the shock to mortality in 2020 and 2021, we expect that it will take several years of new data to establish the post pandemic mortality improvement trend. In the meantime, because we produce regular analysis of pricing for both annuities and longevity swaps, we have a clear view of the market consensus for longevity improvement. We also regularly poll our clients and prospects to get an idea of any likely changes in view.

Whilst we take account of the wider consensus when forming our views, we are not bound to it, as evidenced by our calling out of the “market dislocation” in 2016 (where emerging data suggested mortality improvements had slowed).